Research Article: Longitudinal changes in the relative toxicity of FDA-approved oncology therapeutics: evidence from paired initial and updated RCT reports

Abstract:
A growing share of oncology drugs receive FDA approval via expedited pathways based on interim RCT analyses, leaving safety data relatively immature at approval. Whether the relative toxicity of approved therapeutics shifts as trials mature remains unquantified. Phase II–III RCTs underpinning FDA oncology approvals from 2006 to 2025 were eligible if both the approval-stage dataset and at least one subsequent publication reported grade ?3 AEs or serious AEs (SAEs). For each trial, experimental-versus-control odds ratios (ORs) were computed at both timepoints, with within-trial change indexed by the ratio of odds ratios (ROR), defined as the updated OR divided by the initial OR (ROR?>?1 indicating greater experimental-arm toxicity after maturation). Random-effects meta-analyses pooled log-scale estimates; meta-regression explored effect modifiers and sources of heterogeneity. Ninety-nine RCTs met eligibility and, after splitting seven three-arm trials, yielded 106 pairwise comparisons (103 for grade ?3 AEs; 70 for SAEs). Pooled OR for grade ?3 AEs rose from 1.09 (95% CI, 0.93–1.22) to 1.29 (1.03–1.43); meta-analytic ROR showed a statistically significant increase at 1.095 (1.038–1.156; p =?0.001; I 2 =?12.8%), with 78/103 comparisons shifting toward greater experimental toxicity. For SAEs, pooled OR rose from 1.32 (1.21–1.42) to 1.45 (1.27–1.59); ROR was 1.06 (0.97–1.17; p =?0.22; I 2 =?46.8%), directionally concordant but not statistically significant, with upward shifts in 49/70 comparisons. Amplification was most evident for all-cause events and in crossover-permitted, blinded, medium-sized, and later-line trials, and in lung, prostate, and breast cancer. RORs declined across more recent cohorts ( ? =??0.027, p =?0.009 for SAEs); larger enrollment, longer follow-up, blinded designs, and all-cause ascertainment predicted greater drift, while mature initial follow-up and metastatic settings attenuated it. We derived an exploratory Toxicity Drift Predictor to estimate post-approval ROR of SAEs from pre-approval trial features. FDA-approved oncology therapeutics show systematic post-approval increases in relative toxicity, most pronounced for all-cause events and designs prolonging differential exposure. Though attenuated for recent agents, the approval-stage safety profile represents a lower-bound estimate of long-term comparative harm, supporting evidence-maturity–based dynamic oversight with predefined safety-update milestones and structured surveillance of delayed toxicities.

Introduction:
A growing share of oncology drugs receive FDA approval via expedited pathways based on interim RCT analyses, leaving safety data relatively immature at approval. Whether the relative toxicity of approved therapeutics shifts as trials mature remains unquantified.

Read more